1. Field of the Invention
This invention relates to a semiconductor memory device such as a flash memory and more particularly to a method which is used for detecting the result of verify readout in the programming operation and erase operation and is adequately applied to a flash memory having a large batch processing unit.
2. Description of the Related Art
In a flash memory, the effective programming speed and erase speed are enhanced by programming data in the unit of page and performing the erase operation in the unit of page or in the unit of plural pages. In the programming operation in the page unit, a program pulse applying operation and verify readout operation are performed after program data is serially input and data is input to registers of one page, and then the program pulse applying operation and verify readout operation are repeatedly performed until whole data of the page is programmed. At this time, two methods are conventionally used to determine whether whole data in the page has been programmed or not.
The first method is to detect data of the verify readout result from the all of the registers of the page since the verify readout result is latched in the registers in one page after the verify readout operation was effected after the program pulse applying operation. The first method is schematically explained with reference to FIG. 1.
FIG. 1 is a block diagram showing a core section of a NAND flash memory. In the core section, a memory cell array 100, input/output buffer 110, row decoder 120 and the like are provided. The memory cell array 100 uses different I/O at the time of data input/output and is divided into a plurality of memory cell regions 100-0, 100-1, . . . , 100-i (well regions are not divided). For example, the memory cell array is divided into regions 100-0 to 100-511 when one page is constructed by 512 columns. In FIG. 1, for example, the memory cell 100 which is constructed by eight I/Os is shown and data of one page is supplied via column selection gates controlled by column selection signals CSL0, CSL1, . . . , CSLi and input/output via the input/output buffer 110 (I/O0 to I/O7) for each byte (for each register group 10a of one byte).
Each NAND memory cell 130 is configured by two selection transistors S1, S2 whose gates are respectively connected to selection gate lines SGS, SGD and memory cells MC0, MC1, . . . , MC15 whose current paths are serially connected between the selection transistors S1 and S2 and whose gates are respectively connected to word lines WL0, WL1, . . . , WL15. One end of the current path of the selection transistor S1 is connected to a source line CELSRC and one end of the current path of the selection transistor S2 is connected to a corresponding one of bit lines BL0_0 to BL0_7. The selection gate lines SGS, SGD and word lines WL0, WL1, . . . , WL15 are supplied with predetermined voltage from the row decoder 120.
The bit lines BL0_0 to BL0_7 are respectively connected to registers (page buffers) P/B_0 to P/B_7 which temporarily hold program data. The registers P/B_0 to P/B_7 respectively correspond to the data input/output buffers I/O0 to I/O7 and correspond to the same address in the register group 10a of one byte. Since registers P/B which are connected to other bit lines BLi_0 to BLi_7 in the same manner as described above are provided, registers of the same number are connected for memory cells of one page which are dealt with as a programming unit (i.e. 512 Byte).
Each of the registers P/B_0 to P/B_7 is used for performing both of the operations for holding program data and holding readout data and the verify readout result is held in each of the registers P/B_0 to P/B_7. Therefore, information on whether whole data in one page could be programmed or not can be acquired by reading out and supplying data of the registers P/B_0 to P/B_7 to signal lines DIO0 to DIO7 and directly counting a data number (which is hereinafter simply referred to as a fail number) corresponding to the number of programming fails. In the first method, time for reading out whole data of the registers in one page is required, but it is possible to specifically acquire information on the fail number by use of a counter circuit (not shown) or the like.
On the other hand, the second method is to detect output nodes of registers in one page based on the OR logic. This example is explained with reference to FIG. 2. In the example of FIG. 2, signal lines (nodes) COM0, COM1, . . . , COMi used for outputting data after the verify readout operation are led out from the register group 10a of one byte and respectively connected to the gates of PMOS (P-channel MOS) transistors 22-0 to 22-i and the PMOS transistors 22-0 to 22-i are connected together via fuse elements (Fuse) 21-0 to 21-i in a wired OR configuration. The current path of an NMOS (N-channel MOS) transistor 300 is connected between a common node LSEN of the fuse elements 21-0 to 21-i and the ground node and the gate thereof is controlled by a detection signal VERIFY. Before the detection operation, the signal VERIFY is controlled and set at the xe2x80x9cHxe2x80x9d level and the common node LSEN is set at an xe2x80x9cLxe2x80x9d level.
The fuse elements 21-0 to 21-i are provided to separate a defective column which cannot be used for correctly programming data from the detection operation. In a case where the signal VERIFY is set to the xe2x80x9cLxe2x80x9d level, the register group 10a of each column is connected to the signal lines COM0 to COMi of each column at preset timing and an xe2x80x9cLxe2x80x9d level is output to a signal line COMj (0xe2x89xa6jxe2x89xa6i) of a certain column, then the common node LSEN is charged from the xe2x80x9cLxe2x80x9d level set in the initial state to the xe2x80x9cHxe2x80x9d level via the PMOS transistor 22-j and fuse element 21-j. Since current passage circuits respectively formed by serially connecting the PMOS transistors 22-0 to 22-i and the fuse elements 21-0 to 21-i are connected in parallel, the common node LSEN cannot be maintained at the xe2x80x9cLxe2x80x9d level if the signal lines COM0 to COMi of all of the columns do not output the xe2x80x9cHxe2x80x9d level. Therefore, whether or not pass results of verify readout are held in all of the registers can be detected by making such a construction that data of the register 10b which holds the fail result is output to the signal line COMj as the xe2x80x9cLxe2x80x9d level. With the above method, a period required for outputting data from the register group 10a to the signal lines COM0 to COMi and a period required for determining the potential of the common node LSEN can be made short and the presence or absence of the program fail state can be detected by the batch processing operation.
However, in the case of the first method, it takes a long time to acquire the result, and in the case of the second method, only information that all of the columns in the page have passed can be acquired.
When the fail number detecting operation is performed in the erase operation or during the normal program operation, the operation speed of the second method is high and it is functionally sufficient. However, when a defective column is contained, the second method can be functionally used only after cutting off a fuse element among the fuse elements 21-0 to 21-i which corresponds to the defective column. This is because a defective column section continuously holds fail data in some cases. Therefore, a test process which can be performed only after cutting off the fuse element is provided, the efficiency is lowered.
Further, the state may be permitted as the pass state in some cases if an error correction function is provided inside or outside the semiconductor memory device and the number of fails occurring in the program operation or erase operation is smaller than a preset number within a range of the error correction processing ability. In this case, it is possible to detect a preset number of fails by use of the first method, but it is required to serially read out data after the verify readout operation from all of the registers in the page and there occurs a problem that it takes a long time.
Therefore, it is desired to determine whether the verify readout results are all passed or not and develop a method for detecting the fail number at high speed.
According to an aspect of the present invention, there is a provided a semiconductor memory device comprising: a memory cell array including electrically rewritable nonvolatile memory cells arranged therein; a plurality of latch circuits which temporarily hold data read out from the memory cell array; a first circuit configured to generate a first current which varies in proportion to xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of one end of the plurality of latch circuits; a second circuit configured to generate a predetermined second current; a current control circuit connected to the first and second circuits, and configured to determine absolute values of the first current and the second current, and a third circuit configured to compare the first current and the second current, wherein the number of xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of the one end of the plurality of latch circuit is detected based on a result of comparison between the first current and the second current.
According to another aspect of the present invention, there is provided a semiconductor memory device comprising: a memory cell array including electrically rewritable nonvolatile memory cells arranged therein; a plurality of latch circuits which temporarily hold data to be written into the memory cell array, a first circuit configured to generate a first current which varies in proportion to xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of one end of the plurality of latch circuits; a second circuit configured to generate a predetermined second current, a current control circuit connected to the first and second circuits, and configured to determine absolute values of the first current and the second current; and a third circuit configured to compare the first current and the second current, wherein the number of xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of the one end of the plurality of latch circuit is detected based on a result of comparison between the first current and the second current.
According to still another aspect of the present invention, there is provided a semiconductor memory device comprising: a memory cell array including electrically rewritable nonvolatile memory cells arranged therein; a plurality of latch circuit groups each having a predetermined number of latch circuits which temporarily hold data read out from the memory cell array or data to be written into the memory cell; a first circuit configured to generate a first current which varies in proportion to xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of one end of the plurality of latch groups; a second circuit configured to generate a predetermined second current; a current control circuit connected to the first and second circuits, and configured to determine absolute values of the first current and the second current, wherein the number of xe2x80x9c1xe2x80x9d or xe2x80x9c0xe2x80x9d of binary logical data of one end of the plurality of latch circuits is detected based on a result of comparison between the first current and the second current.